We are facing problems in transcription if numerals and alphabets are written slightly close to one another.
For e.g. if we write something like
it gets transcribed to some
Is there a way to prevent this?
Thank you for your question.
Currently, it seems you did not attach the returned result.
Indeed, maybe the expected result was not added in your lexicon?
If the lexicon doesn't help, then it is more related to a user habit, for which we cannot do much...
Hi Olivier, I had used the below in my conf file. In my resources file I had added the following 20 ml 30 mg When I try to enter the following 20 ml ( the transcription works ) but as soon as I enter 30 mg in the next line, the transcription changes to "20 me 30 mg" Please see the attached snapshot.
Thank you for the precision.
Currently, this use-case is difficult to solve by itself.
We are in a "limit" use-case. If only posology is to be recognized, maybe you should try removing the en_US-lk-text.res resource and use a GRM instead. Otherwise, it lk-text is needed, not much can be done.